RESEARCH · RESEARCH · #1071
Nvidia's SoL-Pi automates harness optimization, cutting coding-agent token use nearly in half
Nvidia researchers introduce SoL-Pi, a system that automatically proposes and tests changes to the control layer (harness) of coding agents. On the EdgeBench tasks the most efficient SoL-Pi variant used 44.7–49% fewer tokens while reaching about 93.7% of Pi harness performance; the system was optimized on GPT-5.6 Sol and also applied to Opus 5 with similar cost savings but reduced mechanism activation, and the paper estimates hourly API savings versus native Codex/Claude Code and versus Pi.
KEY POINTS
- Nvidia researchers introduce SoL-Pi, a system that automatically proposes and tests changes to the control layer (harness) of coding agents.
- On the EdgeBench tasks the most efficient SoL-Pi variant used 44.7–49% fewer tokens while reaching about 93.7% of Pi harness performance; the system was optimized on GPT-5.6 Sol and also applied to Opus 5 with similar cost savings but reduced mechanism activation, and the paper estimates hourly API savings versus native Codex/Claude Code and versus Pi.
- Reducing token consumption at the harness level by roughly half can materially lower runtime costs for long-running coding agents and enables automated harness tuning across models, though the paper notes risks of overfitting and mixed generalization.
WHY IT MATTERS
Reducing token consumption at the harness level by roughly half can materially lower runtime costs for long-running coding agents and enables automated harness tuning across models, though the paper notes risks of overfitting and mixed generalization.